Two experiments replicated and extended research by Croyle and Cooper (1983) indicating that cognitive dissonance involves physiological arousal. In Experiment 1, subjects wrote counterattitudinal essays under conditions of high or low choice, and, to assess arousal effects owing to effort, with or without a list of arguments provided by the experimenter. In high-choice conditions only, and regardless of effort, subjects showed both arousal (heightened galvanic skin response) and attitude change. Arousal, however, did not decline following attitude change. The more effortful task (no arguments provided) produced increased arousal but not greater attitude change. In Experiment 2, the opportunity to change one's attitude following a freely chosen counterattitudinal essay was manipulated. As in Experiment 1, arousal increased following the essay but did not decline following a postessay attitude change opportunity. When subjects were not given an attitude change opportunity, however, arousal did decline. Thus, dissonance seems to create arousal, but attitude change sustains rather than reduces the arousal. It is suggested that if dissonance is a drive state, drive reduction typically may be accomplished through gradual cognitive change or forgetting.